This is a repository copy of Diversity and composition of tropical butterflies along an Afromontane agricultural gradient in the Jimma Highlands, Ethiopia.
White Rose Research Online URL for this paper: http://eprints.whiterose.ac.uk/129734/
Version: Accepted Version
Article: Norfolk, Olivia, Asale, Abebe, Temesgen, Tsegab et al. (4 more authors) (2017) Diversity and composition of tropical butterflies along an Afromontane agricultural gradient in the Jimma Highlands, Ethiopia. Biotropica. pp. 346-354. ISSN 0006-3606 https://doi.org/10.1111/btp.12421
Reuse Items deposited in White Rose Research Online are protected by copyright, with all rights reserved unless indicated otherwise. They may be downloaded and/or printed for private study, or other acts as permitted by national copyright laws. The publisher or other rights holders may allow further reproduction and re-use of the full text version. This is indicated by the licence information on the White Rose Research Online record for the item.
Takedown If you consider content in White Rose Research Online to be in breach of UK law, please notify us by emailing [email protected] including the URL of the record and the reason for the withdrawal request.
[email protected] https://eprints.whiterose.ac.uk/ 1 Norfolk et al. Butterfly diversity along an agricultural gradient
2 Diversity and composition of tropical butterflies along an Afromontane
3 agricultural gradient in the Jimma Highlands, Ethiopia
4 Olivia Norfolk1,2, Abebe Asale3, Tsegab Temesgen3, Dereje Denu3, Philip J. Platts4, Rob
5 Marchant1 and Delenasaw Yewhalaw5,6
6 1 York Institute for Tropical Ecosystems, Environment Department, University of York,
7 Heslington, York, UK, YO10 5NG
8 2 Department of Life Sciences, Anglia Ruskin University, Cambridge, UK, CB1 1PT
9 3 Department of Biology, College of Natural Sciences, Jimma University, Jimma, Ethiopia
10 4 Department of Biology, University of York, Heslington, York, UK, YO10 5DD
11 5Department of Medical Laboratory Sciences and Pathology, College of Health Sciences,
12 Jimma University, Jimma, Ethiopia
13 6Tropical and Infectious Diseases Research Center, Jimma University, Jimma, Ethiopia
1
1 ABSTRACT
2 Afromontane landscapes are typically characterised by a mosaic of smallholder farms and the
3 biodiversity impacts of these practices will vary in accordance to local management and
4 landscape context. Here we assess how tropical butterfly diversity is maintained across an
5 agricultural landscape in the Jimma Highlands of Ethiopia. We used transect surveys to
6 sample understory butterfly communities within degraded natural forest, semi-managed
7 coffee forest (SMCF), exotic timber plantations, open woodland, croplands and pasture.
8 Surveys were conducted in 29 one-hectare plots and repeated five times between January and
9 June 2013. We found that natural forest supports higher butterfly diversity than all
10 agricultural plots (measured with Hill’s numbers). SMCF and timber plantations retain
11 relatively high abundance and diversity, but these metrics drop off sharply in open woodland,
12 cropland and pasture. SMCF and timber plantations share the majority of their species with
13 natural forest and support an equivalent abundance of forest-dependent species, with no
14 increase in widespread species. There was some incongruence in the responses of families
15 and sub-families, notably that Lycaenidae are strongly associated with open woodland and
16 pasture. Adult butterflies clearly utilise forested agricultural practices such as SMCF and
17 timber plantations, but species diversity declines steeply with distance from natural forest
18 suggesting that earlier life-stages may depend on host plants and/or microclimatic conditions
19 that are lost under agricultural management. From a management perspective, the protection
20 of natural forest remains a priority for tropical butterfly conservation, but understanding
21 functioning of the wider landscape mosaic is important as SMCF and timber plantations may
22 act as habitat corridors that facilitate movement between forest fragments.
23 Keywords: Africa; agroforestry; cropland; coffee; Ethiopia; farming; land-use change;
24 tropical forest
2
1 TROPICAL DEFORESTATION IS A MAJOR DRIVER OF BIODIVERSITY DECLINES (Dirzo & Raven
2 2003, Gaston et al. 2003), one which continues at pace in response to anthropological
3 pressures such as increasing food and timber demands (Geist & Lambin 2002, Lawrence &
4 Vandecar 2015, Lewis et al. 2015). Expanding production landscapes are unlikely to match
5 the conservation value of natural forests, but many traditional agricultural systems can
6 provide an important refuge for biodiversity (Torquebiau 1992, Bhagwat et al. 2008, Jose
7 2009). Afromontane landscapes tend to incorporate a broad range of agricultural systems,
8 ranging from traditional agroforestry systems, mono-culture timber plantations, mixed
9 croplands to pasture. Understanding the extent to which these different agricultural systems
10 contribute towards the maintenance of tropical biodiversity will help inform future landscape
11 management and may facilitate the development of nature-based strategies that enhance food
12 production whilst maintaining biodiversity and ecosystem services (Fischer et al. 2014).
13 Tropical butterflies are a highly diverse group of organisms (Bonebrake et al. 2010) that due
14 to short generation times and high mobility tend to exhibit high sensitivity to land-use change
15 (Lawton et al. 1998, Thomas et al. 2001). In general, butterfly diversity tends to decrease
16 when tropical forest is converted into agricultural land, but the magnitude of this effect can
17 differ quite considerably between agricultural systems. Studies in Indonesia and South
18 America have found that agricultural landscapes support reduced butterfly species richness
19 when compared to tropical forest, but that agroforestry systems support higher numbers of
20 species than annual cropland and pasture (Schulze et al. 2004, Barlow et al. 2007,
21 Francesconi et al. 2013). In Western Africa, cashew forest plantations have been linked to a
22 reduction in butterfly species richness (Vasconcelos et al. 2015) as have annual cultures,
23 though Cameroonian agroforests can support species richness equal to natural forest (Bobo et
24 al. 2006). The conservation value of agricultural systems can also vary in accordance with
25 landscape context and tends to decrease with isolation from natural forest (Horner Devine et
‐ 3
1 al. 2003, Schulze et al. 2004, Munyuli 2013). In Costa Rica, plots incorporating both
2 agroforestry and natural forest supported higher species richness than forest plots alone
3 (Horner Devine et al. 2003), further emphasising the importance of assessing agricultural
4 impacts ‐in the context of the wider landscape.
5 In Ethiopia, forest cover has declined from 40% to 2.7% since the beginning of the
6 20th century, primarily as a result of expanding agricultural pressures (Pohjonen & Pukkala
7 1990). The Ethiopian Highlands are a major component of the Eastern Afromontane
8 Biodiversity Hotspot (Mittermeier et al. 2004), covering half of its area, yet the impacts of
9 agricultural expansion on biodiversity remain relatively understudied. Highland communities
10 such as the Jimma Highlands (Fig S1) have a long history of coffee production, with wild
11 coffee traditionally harvested from natural forests. To increase yields coffee growers modify
12 natural forests by thinning trees and removing lianas and shrubs. These semi-managed coffee
13 forests (SMCFs) form a characteristic feature of the Jimma Highlands and the retention of
14 canopy trees means that they are likely to play a valuable role in the conservation of forest-
15 dependent wildlife, especially when compared with more intensive forms of land use. Indeed
16 Ethiopian shaded coffee has been shown to support high levels of bird diversity (Buechley et
17 al. 2015) and may have similar benefits for tropical butterflies in the region.
18 The Jimma Highlands also support exotic timber plantations which have expanded
19 five-fold in the past 20 years due to increasing demands for timber and firewood (Bekele
20 2011). These monoculture plantations do not retain native forest trees but do support some
21 ground flora and understory species associated in un-cleared forest. At lower elevations
22 agricultural practices become more intensive and are typified by croplands of annual cultivars
23 and pasture. This study assesses the relative value of these agricultural systems in terms of
24 tropical butterfly conservation, comparing butterfly abundance, diversity and composition
25 along a land-use gradient ranging from natural forest to SMCF, timber plantation, open
4
1 woodland, cropland and pasture. We consider how local land use, tree diversity and
2 landscape factors influence butterfly communities, allowing us to assess the relative
3 contribution that different agricultural practices make towards butterfly conservation in the
4 region.
5 METHODS
6 STUDY SITE. - The study took place in the Jimma Highlands of south-western Ethiopia, along
7 a 20 km transect running between the Gumay and Setema Districts. The study transect
8 spanned an altitudinal gradient of 1500 m to 2226 m and incorporated a range of land-use
9 types that are representative of the region, from the intensively managed pasture and cropland
10 associated with the lowlands, up towards the lesser disturbed forest of the highlands (Fig S1).
11 Land use was classified into six distinct categories: natural forest, semi-managed coffee
12 forest (SMCF), timber plantation, open woodland, cropland and pasture. The characteristics
13 of these land-use categories are defined in Table 1.
14 BUTTERFLY SURVEYS. -We conducted butterfly surveys in 29 × 1 ha plots (Fig. S1),
15 encompassing natural forest (4 plots), SMCF (7), timber plantation (3), open woodland (4),
16 cropland (6) and pasture (5). We selected plots through a stratified random sampling design,
17 whereby we identified the main land-use types for the transect using 2008 SPOT5 satellite
18 imagery (Hailu et al., 2014) and placed the 1 ha plots randomly in each land-use type. We
19 surveyed each plot five times between January and June 2015, a period that encompassed the
20 end of the dry season and the beginning of the rainy season (survey one, 31 December 2014 –
21 9 January 2015; survey two, 26 January – 5 February; survey 3, 28 March – 6 April; survey
22 four, 2–11 May; survey five, 1-10 June).
23 Within each plot, we recorded butterflies along five 50 m line transects, spaced at 25 m
24 intervals and traversed in alternate directions. We walked transects at a steady pace,
5
1 recording all butterflies observed within 2.5 m either side of the transects and 5 m vertically.
2 When possible we photographed butterflies during the transect counts to aid identification.
3 The majority of individuals were identified to species level, but when species level
4 identification was not possible, butterflies were classified into morpho-species. Surveys were
5 all conducted between 0900 h and 1630 h on sunny, windless days. Data collected from the
6 five transects were pooled per plot. Species were assigned to ecological habitat categories in
7 accordance with Munyuli (2012) (nomenclature adapted from Larsen 1996): FDS, forest
8 dependent species; FEW, forest edge and woodland species; MS, migratory species; OHPS,
9 open habitat specialists; or WSS, widespread species.
10 ENVIRONMENTAL VARIABLES.-We conducted tree surveys in all 29 plots in April 2014. We
11 identified to species level all woody stems with a diameter at breast height (dbh) ≥ 10 cm.
12 Using these data we calculated stem density and tree species richness per 1 ha plot. We also
13 surveyed herbaceous plants and shrubs in five 1 m × 1 m quadrats that were randomly
14 distributed within each plot, identifying all individuals to species level and then collated these
15 data per plot.
16 In order to consider the effect of isolation from natural forest, we estimated linear distance
17 from each sampling point to the nearest patch of natural forest using land cover data that was
18 created using a supervised classification of SPOT satellite imagery for the year 2008 (Hailu
19 et al., 2014). Plots were also categorised into five altitudinal bands for analyses: 1) 1500-
20 1636 m; 2) 1637-1779 m; 3) 1780-1836 m; 4) 1837- 2089 m and 5) 2090- 2226 m.
21 STATISTICAL ANALYSIS.- Statistical analyses were conducted in R version 3.2 (R Core Team,
22 2015) using the vegan package (Oksanen et al. 2012). Species accumulation curves were
23 created for each land-use type. Alpha diversity was calculated for each point count using
24 Hill’s numbers (Hill 1973). Hill's numbers are defined to the order of q (qD), whereby
6
1 parameter q indicates the weight given towards rare or common species. 0D (species richness)
2 is insensitive to relative frequencies, and is therefore weighted towards rare species, 1D
3 (exponential of Shannon) is weighted towards common species, and 2D (inverse Simpson) is
4 weighted towards abundant species. These diversity indices are particularly useful because
5 they are scalable and can provide insight into the representation of rare, common and
6 abundant species within different land-use types (Jost 2006, Tuomisto 2010, Chao et al.
7 2012).
8 Pair-wise species similarity was calculated between natural forest and the five other
9 land-use types (Forest-Plantation, Forest-SMCF, Forest-Woodland, Forest-Cropland, Forest-
10 Pasture). Species similarity was also weighted by the aforementioned q to provide insight into
11 the relative abundance of those shared species; q=0 was calculated as the Sorenson similarity
12 index (insensitive to species abundance), q=1 as the Horn index (weighted towards common
13 species) and q=2 as the Morisita index (weighted towards abundant species) (Chao et al.
14 2012). This combination of metrics provides insight into not only the proportion of species
15 shared, but the relative abundances of those shared species
16 Linear mixed-effect models were used to assess the impact of land use and
17 environmental variables on butterfly abundance and all three measures of Hill’s diversity
18 using the lme4 package (Bates 2005). Response variables were log-transformed to normalise
19 the data. The fixed effects included in the full models were: 1) management type, 2) distance
20 from nearest patch of natural forest (considered zero for plots within natural forest) and 3)
21 vegetation (tree density, tree species richness, herb species richness, shrub species richness).
22 Initial investigation suggested that there was considerable seasonal variation in abundance
23 and species richness. To account for this temporal variation in the replicated plots, we
24 included survey round as a random intercept. We also included altitudinal zone as a random
25 intercept to account for spatial autocorrelation of plots along the altitudinal gradient. Best
7
1 fitting models were selected using the dredge function in R, which returned models with the
2 lowest AIC values (delta AIC < 4). The strength of the fixed effects retained in the best
3 fitting models were assessed using marginal R2 values calculated using the MuMIn package
4 (Barton 2014) and their significance was determined by comparing the fit of subsequent
5 models using Chi-squared tests (Zuur et al. 2009). Equivalent models were also run for
6 butterfly abundance within the five ecological habitat categories (FDS, FEW, MS, OHPS and
7 WSS), and within the six most abundant sub-families (Coliadinae, Pierinae, Satyrinae,
8 Heliconiinae, Lycaeninae and Papilioninae).
9 Non-metric multidimensional scaling (NDMS) was used to assess how community
10 composition was affected by land use. This unconstrained ordination technique collapses the
11 species data into two dimensions, allowing differences between land-use categories to be
12 visualised. Because it relies upon rank-orders (rather than absolute abundance) it can
13 accommodate non-linear species responses, allowing the detection of underlying responses to
14 environmental change (Oksanen et al. 2012). The significance of land use was assessed using
15 permutation tests (999 permutations) with the envfit function in R.
16 RESULTS
17 A total of 6616 butterflies were recorded, belonging to 64 species (and six morpho-species),
18 the majority of which were fruit-feeding butterflies from the family Nymphalidae (44),
19 followed by Pieridae (19), Papilionidae (5) and Lycaenidae (2) (Table S1 for full species list).
20 Species accumulation curves had not reached their asymptotes, but there was clear separation
21 between land-use types, with natural forest, timber plantation and SMCF exhibiting steeper
22 rates of accumulation than open woodland, pasture and cropland (Fig. 1a). Estimated species
23 richness was highest within timber plantations (Chao ± SE: 79 ± 17), followed by SMCF (72
24 ± 9) and forest (64 ± 3). Estimated species richness was similar in open woodland (48 ± 6),
8
1 pasture (51 ± 12) and cropland (49 ± 11). Of the 70 recorded species, three nymphalid species
2 were unique to natural forest (Précis octavia, Charaxes karkloof and Acraea cerasa), two to
3 SMCF (Acraea alciope and Junonia natalica), one to woodland (Pseudacraea eurytus) and
4 one to pasture (Junonia hierta). Timber plantation and cropland did not contain any unique
5 species. The most numerous species overall was Colias electo (16% of all individuals), which
6 was found to be most abundant in natural forest, SMCF and plantation.
7 Butterfly abundance per plot differed significantly between land-use types (Table 2)
8 and was highest in SMCF (Individuals per ha ± SE: 41 ± 5), natural forest (37 ± 6) and
9 plantations (35 ±5). Open woodland supported intermediate levels of abundance (23 ± 6), but
10 numbers dropped sharply in pasture (10 ± 3) and cropland (6 ± 1). Hill’s diversity per plot
11 also differed significantly with land use at all levels of q (Table 2). Natural forest supported
12 the highest levels of butterfly diversity (Fig. 1b) followed by plantation and SMCF. Open
13 woodland supported intermediate levels of diversity, but pasture and cropland supported less
14 than a quarter of the diversity associated with natural forest, SMCF and plantation. These
15 trends were true at all levels of q, indicating higher numbers of rare, common and abundant
16 species in the forested habitats. Diversity decreased steeply to the order of q in forest,
17 plantations and SMCF indicating that high numbers of species occurred at low abundances,
18 with fewer species common or abundant. Within pasture and cropland, diversity showed little
19 decline to the order of q, indicating similar numbers of rare, common and abundant species.
20 Butterfly communities associated with timber plantation and SMFC exhibited high
21 levels of species similarity with natural forest communities (Fig. 2). Similarity to forest was
22 high for all orders of q (>80% of species shared), suggesting that not only are timber
23 plantations and SMCF supporting similar species to those in the forest, but that those species
24 are occurring at similar relative abundance. Open woodland, cropland and pasture showed
25 much lower levels of similarity to natural forest communities. These habitats all exhibited a
9
1 sharp drop in similarity between q=0 and q=1 suggesting that although approximately 60% of
2 forest species were present in open woodland, pasture and cropland, the identities of common
3 and abundant species differed considerably from those associated with natural forest.
4 In addition to land use, the best-fitting models included distance from the nearest
5 patch of natural forest, with butterfly abundance and diversity (at all orders of q) declining
6 with distance (Table 2). At distances of 500 m diversity was approximately half of that
7 associated with plots adjacent to natural forest, with diversity halving again by 1000 m (Fig.
8 3). Vegetative variables (tree density and tree, herb and liana species richness) explained little
9 variation and were not retained in the final models.
10 COMMUNITY COMPOSITION. - Forty-two of the 70 recorded butterfly species were assigned to
11 an ecological habitat category (Table S1), with the categorised species making up 78% of all
12 observed individuals. The majority of butterflies were migratory species (57% of categorised
13 observations), followed by open habitat specialists (20%) and widespread species (13%).
14 Forest-dependent species and forest edge/woodland species made up just 7% and 1% of
15 observations respectively. All ecological habitat categories exhibited a significant response to
16 land use (Table S3), but the strength and direction of the trends differed between groups.
17 Migratory and forest-dependent species showed the strongest responses to land use (Fig. 4a &
18 c). Both groups occurred at similar abundance in natural forest, timber plantation and SMCF,
19 with numbers dropping off sharply in the other land-use types. Forest edge/woodland species
20 showed similar patterns, but trends were less pronounced (Fig. 4b). Open habitat specialists
21 occurred in the lowest numbers within cropland and pasture and were most abundant within
22 timber plantation (Fig. 4d). Widespread species showed a strong preference for open
23 woodland (Fig. 4e).
10
1 All of the common families and sub-families were significantly affected by land use (Table
2 S2), but again the strength and direction of the effect differed between groups (Fig. S2).
3 Within the Pieridae, the abundance of Coliadinae was strongly influenced by land use, with
4 butterflies occurring at high numbers within natural forest, timber plantation and SMCF, and
5 declining sharply in open woodland, pasture and cropland (Fig. S2a). Pierinae showed a
6 weaker response, but exhibited similar trends (Fig. S2b). The Nymphalidae also tended to
7 occur at low abundance within open woodland, cropland and pasture, though trends differed
8 between sub-families; butterflies from Satyrinae were most abundant within plantations (Fig.
9 S2c-d), whereas those from Heliconiinae occurred at low numbers within plantations and
10 were most abundant in natural forest. Papilonidae exhibited similar trends and were most
11 abundant in forest, followed by plantations and SMCF (Fig. S2e). In contrast to the other sub-
12 families, Lycaeninae occurred in low numbers in natural forest, timber plantation, SMCF and
13 cropland, but were highly abundant in open woodland habitats, with intermediate numbers
14 observed in pasture (Fig. S2f).
15 NDMS ordination showed that butterfly communities overlapped considerably
16 between all land-use types, with no significant separation between the land-use categories
17 (Fig. S3: R2=0.036, P=0.396). The ordination did reveal some differences in the composition
18 of butterfly families, with species from Lycaenidae showing strong positive loadings with
19 NDMS axis-1 in association with open woodland and pasture habitats.
20 DISCUSSION
21 Butterfly communities in the Jimma Highlands are strongly influenced by agricultural land
22 use, with both abundance and species diversity decreasing sharply in non-wooded farmland
23 such as cropland and pasture. Natural forest supports the highest level of butterfly diversity
24 per plot, but estimates of species richness across all plots suggests that semi-managed coffee
11
1 forests (SMCFs) and timber plantations support a similar number of species as natural forest,
2 perhaps due to turnover of species between plots. Though SMCFs and plantations were
3 utilised by adult forest butterflies, we found that both abundance and diversity declined with
4 distance from natural forest. This suggests that the persistence of forest species may be
5 contingent on larval host plants or microclimatic conditions present only in the natural forest.
6 Shaded coffee systems are frequently associated with positive biodiversity benefits,
7 outperforming sun coffee farms in terms of butterfly species richness (Perfecto et al. 2003),
8 bird abundance (Komar 2006) and subsequent avian ecosystem services such as pest control
9 (Perfecto et al. 2004, Kellermann et al. 2008). Other studies have even found that coffee
10 agroforest can support higher butterfly species richness than natural forest (Bobo et al. 2006).
11 In the Jimma Highlands, we found that SMCF and timber plantations are utilised by equally
12 diverse butterfly communities, despite the considerable reduction of tree species diversity
13 within plantations. Exotic timber plantations tend to be considered in a less positive light
14 from a conservation perspective, but studies in Brazil have found that although butterfly
15 diversity decreases from natural forest into Eucalyptus plantations (Barlow et al. 2007), the
16 plantations do support a relatively diverse community that benefit from the species-rich
17 understory vegetation. Korean pine plantations have even been shown to maintain butterfly
18 species richness at levels equivalent to natural forest (Lee et al. 2014). Our results confirm
19 that timber plantations are utilised by adult forest butterflies, and at equivalent levels to more
20 diverse agroforestry systems such as SMCF.
21 SPECIES COMPOSITION AND ECOLOGICAL HABITAT CATEGORIES. - Measures of species diversity
22 can be misleading from a conservation perspective, as disturbed forest can often support
23 elevated butterfly species richness as a consequence of increasing numbers of opportunistic
24 and widespread species (Spitzer et al. 1993, Spitzer et al. 1997, Bobo et al. 2006). In our
25 study, species similarity was extremely high between natural forest and SMCF and timber
12
1 plantation (>80% species shared), suggesting that both of these agricultural habitats are being
2 utilised by forest species and not just by opportunistic, widespread species. Consideration of
3 ecological habitat categories confirmed that SMCF and timber plantation support similar
4 numbers of forest-dependent species as natural forest, with no increase in the abundance of
5 widespread species. However, timber plantations do appear to support elevated numbers of
6 open habitat specialists from the sub-family Satyrinnae, a pattern also observed in Brazilian
7 Eucalyptus plantations (Barlow et al. 2007). The Satyrianne exhibit diverse responses to
8 forest disturbance, with some species preferring dense undergrowth (Brown & Freitas 2000,
9 Ghazoul 2002) and others flourishing in the open habitats associated within forest disturbance
10 (Daily & Ehrlich 1995, Shahabuddin & Terborgh 1999). Here the high numbers of open
11 habitat specialists presumably reflects a lack of dense undergrowth within the plantations as
12 compared to natural forest and SMCF.
13 Tropical butterflies can exhibit considerable vertical stratification from ground to
14 canopy level (Molleman et al. 2006, Ribeiro et al. 2015), with canopy assemblages showing a
15 higher susceptibility to decline in disturbed forest or logged forest than those found at ground
16 level (Whitworth et al. 2016, Dumbrell & Hill 2005). Since we used transect surveys
17 focussed on ground-level species, canopy species are likely to be under-represented in our
18 results. Additional sampling of canopy assemblages could reveal a stronger decline from
19 natural forest into SMCF and plantation forest, since the modified tree communities are likely
20 to be associated with changes in canopy structure.
21 PROXIMITY TO NATURAL FOREST. - We recorded a high diversity of adult butterflies in
22 both SMCF and timber plantation, however habitat requirements for butterflies can vary
23 through their life cycle. Larval stages often depend on a specific host plant and require
24 distinct microclimatic conditions from their adult counterparts (Fartmann 2006), and though
25 adult butterflies are frequently observed using forest gaps and edge habitats for basking and
13
1 nectaring (Hill et al. 2001, Tropek & Konvicka 2010, Vlasanek et al. 2013), they typically
2 depend on larval host plants in the understory. An intensive mark-release-recapture study in
3 Papa New Guinea showed that many tropical butterflies can disperse up to 1 km from their
4 larval host plant (Vlasanek et al. 2013), so the presence of adult butterflies within SMCF and
5 timber plantation does not guarantee that these habitats in isolation could support the
6 observed butterfly diversity.
7 This is consistent with our finding that butterfly abundance and diversity both
8 decreased with distance from natural forest, with diversity declining by more than half over
9 distances greater than 500 m. Other studies in tropical agricultural landscapes have found that
10 agricultural land use has a stronger effect on butterfly diversity than proximity to forest
11 (Perfecto et al. 2003, Francesconi et al. 2013), but on a wider scale, isolation from contiguous
12 forest is negatively correlated with the species richness of fruit-feeding butterflies in Bornean
13 forest fragments (Benedick et al. 2006). Landscape effects are known to impact butterfly
14 meta-population dynamics in temperate systems (Hanski & Thomas 1994, Hill et al. 1996,
15 Thomas et al. 2001), with occasional immigration from source populations rescuing isolated
16 populations at marginal ‘sink’ sites from extinction (Hanski & Ovaskainen 2003). The meta-
17 population dynamics of tropical systems are less well-established (Bonebrake et al. 2010),
18 but large areas of forest are likely to act as source populations for more isolated forest
19 fragments. Since SMCF and timber plantations are used by a wide diversity of adult
20 butterflies they may be able to increase the permeability of the agricultural matrix by acting
21 as habitat corridors that enable movement between remaining fragments of natural forest
22 (Haddad & Tewksbury 2005).
23 CONCLUSIONS. - Tropical butterflies are adversely affected by agricultural conversion of
24 natural forest in the Jimma Highlands, but the impact varies dramatically between
25 agricultural practices. Butterfly abundance and diversity are particularly low in non-wooded
14
1 habitats such as cropland and pasture, so the expansion of these agricultural practices would
2 have strong negative implications for butterfly conservation in the region. Semi-managed
3 coffee forests (SMCF) and timber plantations are utilised by a wide variety of forest
4 butterflies, but diversity declines with increasing distance from natural forest suggesting that
5 natural forest remains crucial to the butterfly life-cycle. From a management perspective, the
6 maintenance of natural forest should be a priority for the conservation of forest butterflies,
7 but SMCF and timber plantations may have the potential to act as habitat corridors that
8 facilitate movement of adult butterflies between otherwise isolated forest fragments.
9 ACKNOWLEDGEMENTS
10 Funded by the Ministry for Foreign Affairs of Finland through the CHIESA project
11 (http://chiesa.icipe.org/).
12 Data Availability: The data used in this study are archived at the Dryad Digital Repository
13 (http://dx.doi.org/10.5061/dryad.h6t7g).
15
1 LITERATURE CITED
2 BARLOW, J., I. S. ARAUJO, W. L. OVERAL, T. A. GARDNER, F. DA SILVA MENDES, I. R. LAKE,
3 AND C. A. PERES. 2009. Diversity and composition of fruit-feeding butterflies in
4 tropical Eucalyptus plantations. Plantation Forests and Biodiversity: Oxymoron or
5 Opportunity?, pp. 165-180. Springer.
6 BARLOW, J. O. S., W. L. OVERAL, I. S. ARAUJO, T. A. GARDNER, AND C. A. PERES. 2007. The
7 value of primary, secondary and plantation forests for fruit-feeding butterflies in the
8 Brazilian Amazon. J. Appl. Ecol. 44: 1001-1012.
9 BEKELE, M. 2011. Forest Plantations and Woodlots In Ethiopia: A Platform for Stakeholders
10 in African Forestry. African Forest Forum. Working Paper Series.
11 BENEDICK, S., J. K. HILL, N. MUSTAFFA, V. K. CHEY, M. MARYATI, J. B. SEARLE, M.
12 SCHILTHUIZEN, AND K. C. HAMER. 2006. Impacts of rain forest fragmentation on
13 butterflies in northern Borneo: species richness, turnover and the value of small
14 fragments. J. Appl. Ecol. 43: 967-977.
15 BHAGWAT, S. A., K. J. WILLIS, H. J. B. BIRKS, AND R. J. WHITTAKER. 2008. Agroforestry: a
16 refuge for tropical biodiversity? Trends. Ecol. Evolut. 23: 261-267.
17 BOBO, K. S., M. WALTERT, H. FERMON, J. NJOKAGBOR, AND M. MÜHLENBERG. 2006. From
18 Forest to Farmland: Butterfly Diversity and Habitat Associations Along a Gradient of
19 Forest Conversion in Southwestern Cameroon. J. Insect Conserv. 10: 29-42.
20 BONEBRAKE, T. C., L. C. PONISIO, C. L. BOGGS, AND P. R. EHRLICH. 2010. More than just
21 indicators: A review of tropical butterfly ecology and conservation. Biol. Conserv.
22 143: 1831-1841.
16
1 BROOKS, T. M., R. A. MITTERMEIER, C. G. MITTERMEIER, G. A. B. DA FONSECA, A. B.
2 RYLANDS, W. R. KONSTANT, P. FLICK, J. PILGRIM, S. OLDFIELD, G. MAGIN, AND C.
3 HILTON-TAYLOR. 2002. Habitat Loss and Extinction in the Hotspots of Biodiversity.
4 Conserv. Biol. 16: 909-923.
5 BROWN, K. S., AND A. V. L. FREITAS. 2000. Atlantic Forest Butterflies: Indicators for
6 Landscape Conservation. Biotropica 32: 934-956.
7 BUECHLEY, E. R., Ç. H. ŞEKERCIOĞLU, A. ATICKEM, G. GEBREMICHAEL, J. K. NDUNGU, B. A.
8 MAHAMUED, T. BEYENE, T. MEKONNEN, AND L. LENS. 2015. Importance of Ethiopian
9 shade coffee farms for forest bird conservation. Biol. Conserv. 188: 50-60.
10 DAILY, G.C. AND P.R. EHRLICH. 1995. Preservation of biodiversity in small rain-forest
11 patches – rapid evaluations using butterfly trapping. Biodiv. Conserv. 4: 35–55.
12 DIRZO, R., AND P. H. RAVEN. 2003. Global state of biodiversity and loss. Ann. Rev. Environ.
13 Res. 28: 137-167.
14 DUMBRELL, A. J., AND J. K. HILL. 2005. Impacts of selective logging on canopy and ground
15 assemblages of tropical forest butterflies: Implications for sampling. Biol. Conserv.
16 125: 123-131.
17 FARTMANN, T. 2006. Oviposition preferences, adjacency of old woodland and isolation
18 explain the distribution of the Duke of Burgundy butterfly (Hamearis lucina) in
19 calcareous grasslands in central Germany. Annales Zoologici Fennici. 43: 335-347.
20 FISCHER, J., D. J. ABSON, V. BUTSIC, M. J. CHAPPELL, J. EKROOS, J. HANSPACH, T.
21 KUEMMERLE, H. G. SMITH, AND H. VON WEHRDEN. 2014. Land Sparing Versus Land
22 Sharing: Moving Forward. Conserv. Lett. 7: 149-157.
17
1 FRANCESCONI, W., P. K. R. NAIR, D. LEVEY, J. DANIELS, AND L. CULLEN, JR. 2013. Butterfly
2 distribution in fragmented landscapes containing agroforestry practices in
3 Southeastern Brazil. Agroforest. Syst. 87: 1321-1338.
4 GASTON, K. J., T. M. BLACKBURN, AND K. K. GOLDEWIJK. 2003. Habitat conversion and
5 global avian biodiversity loss. Proc. Roy. Soc. London B: Biol. Sci. 270: 1293-1300.
6 GEIST, H. J., AND E. F. LAMBIN. 2002. Proximate Causes and Underlying Driving Forces of
7 Tropical Deforestation Tropical forests are disappearing as the result of many
8 pressures, both local and regional, acting in various combinations in different
9 geographical locations. BioScience 52: 143-150.
10 GHAZOUL, J. 2002. Impact of logging on the richness and diversity of forest butterflies in a
11 tropical dry forest in Thailand. Biodiv. Conserv. 11: 521-541.
12 HADDAD, N.M. AND J.J. TEWKSBURY. 2005. Low quality habitat corridors as movement
13 conduits for two butterfly species. Ecol. App.‐ 15: 250-257.
14 HAILU, B.T., MAEDA, E.E., HURSKAINEN, P. AND PELLIKKA, P. K. E. 2014. Object-based
15 image analysis for distinguishing indigenous and exotic forests in coffee production
16 areas of Ethiopia. Appl. Geomatics 6: 207-214.
17 HANSKI, I. 1999. Metapopulation Ecology. Oxford University Press, Oxford, UK.
18 HANSKI, I., AND O. OVASKAINEN. 2003. Metapopulation theory for fragmented landscapes.
19 Theor. Popul. Biol. 64: 119-127.
20 HANSKI, I., AND C. D. THOMAS. 1994. Metapopulation dynamics and conservation: a spatially
21 explicit model applied to butterflies. Biol. Conserv. 68: 167-180.
18
1 HILL, J. K., C. D. THOMAS, AND O. T. LEWIS. 1996. Effects of Habitat Patch Size and Isolation
2 on Dispersal by Hesperia comma Butterflies: Implications for Metapopulation
3 Structure. J. Anim. Ecol. 65: 725-735.
4 HILL, J.K., HAMER, K.C., TANGAH, J. AND M. DAWOOD. 2001. Ecology of tropical butterflies
5 in rainforest gaps. Oecologia. 128: 294–302.
6 HORNER DEVINE, M. C., G. C. DAILY, P. R. EHRLICH, AND C. L. BOGGS. 2003. Countryside
7 biogeography‐ of tropical butterflies. Conserv. Biol. 17: 168-177.
8 JOSE, S. 2009. Agroforestry for ecosystem services and environmental benefits: an overview.
9 Agroforest. Syst. 76: 1-10.
10 KELLERMANN, J. L., M. D. JOHNSON, A. M. STERCHO, AND S. C. HACKETT. 2008. Ecological
11 and economic services provided by birds on Jamaican Blue Mountain coffee farms.
12 Conserv. Biol. 22: 1177-1185.
13 KOMAR, O. 2006. Ecology and conservation of birds in coffee plantations of El Salvador,
14 Central America. PhD dissertation. University of Kansas, USA.
15 LARSEN, T.B. 1996. The Butterflies of Kenya and Their Natural History, Oxford University
16 Press, Oxford, UK.
17 LAWRENCE, D., AND K. VANDECAR. 2015. Effects of tropical deforestation on climate and
18 agriculture. Nat. Climate Change 5: 27-36.
19 LAWTON, J. H., D. E. BIGNELL, B. BOLTON, G. F. BLOEMERS, P. EGGLETON, P. M. HAMMOND,
20 M. HODDA, R. D. HOLT, T. B. LARSEN, N. A. MAWDSLEY, N. E. STORK, D. S.
21 SRIVASTAVA, AND A. D. WATT. 1998. Biodiversity inventories, indicator taxa and
22 effects of habitat modification in tropical forest. Nature 391: 72-76.
19
1 LEE, C. M., T.-S. KWON, S.-S. KIM, J.-D. SOHN, AND B.-W. LEE. 2014. Effects of forest
2 degradation on butterfly communities in the Gwangneung Forest. Entomological
3 Science 17: 293-301.
4 LEWIS, S. L., D. P. EDWARDS, AND D. GALBRAITH. 2015. Increasing human dominance of
5 tropical forests. Science 349: 827-832.
6 MOLLEMAN, F., A. KOP, P. BRAKEFIELD, P. DE VRIES, AND B. ZWAAN. 2006. Vertical and
7 Temporal Patterns of Biodiversity of Fruit-Feeding Butterflies in a Tropical Forest in
8 Uganda. Biodiv. Conserv. 15: 107-121.
9 MUNYULI, M.B., 2012. Butterfly diversity from farmlands of central Uganda. Psyche. Article
10 ID: 481509: 1-23.
11 MUNYULI, M. B. T. 2013. Drivers of species richness and abundance of butterflies in coffee–
12 banana agroforests in Uganda. Int. J. Biodivers. Sci. Ecosyst. Serv. Manage. 9: 298-
13 310.
14 MYERS, N., R. A. MITTERMEIER, C. G. MITTERMEIER, G. A. B. DA FONSECA, AND J. KENT.
15 2000. Biodiversity hotspots for conservation priorities. Nature 403: 853-858.
16 PERFECTO, I., A. MAS, T. DIETSCH, AND J. VANDERMEER. 2003. Conservation of biodiversity
17 in coffee agroecosystems: a tri-taxa comparison in southern Mexico. Biodiv. Conserv.
18 12: 1239-1252.
19 PERFECTO, I., J. H. VANDERMEER, G. L. BAUTISTA, G. I. NUNEZ, R. GREENBERG, P. BICHIER,
20 AND S. LANGRIDGE. 2004. Greater predation in shaded coffee farms: the role of
21 resident neotropical birds. Ecol. 85: 2677-2681.
20
1 POHJONEN, V., AND T. PUKKALA. 1990. Eucalyptus globulus in Ethiopian forestry. Forest
2 Ecol. Management 36: 19-31.
3 RIBEIRO, D. B., M. R. WILLIAMS, A. SPECHT, AND A. V. L. FREITAS. 2015. Vertical and
4 temporal variability in the probability of detection of fruit-feeding butterflies and
5 moths (Lepidoptera) in tropical forest. Austral Entomol. 55:112-120.
6 SCHULZE, C., I. STEFFAN-DEWENTER, AND T. TSCHARNTKE. 2004. Effects of Land Use on
7 Butterfly Communities at the Rain Forest Margin: A Case Study from Central
8 Sulawesi. In G. Gerold, M. Fremerey and E. Guhardja (Eds.). Land Use, Nature
9 Conservation and the Stability of Rainforest Margins in Southeast Asia, pp. 281-297.
10 Springer Berlin Heidelberg.
11 SHAHABUDDIN, G. AND J. W. TERBORGH. 1999. Frugivorous butterflies in Venezuelan forest
12 fragments: abundance, diversity and the effects of isolation. J. Trop. Ecol. 15: 703–
13 722.
14 THOMAS, J. A., N. A. D. BOURN, R. T. CLARKE, K. E. STEWART, D. J. SIMCOX, G. S. PEARMAN,
15 R. CURTIS, AND B. GOODGER. 2001. The quality and isolation of habitat patches both
16 determine where butterflies persist in fragmented landscapes. Proc. Roy. Soc. London
17 B: Biol. Sci. 268: 1791-1796.
18 TORQUEBIAU, E. 1992. Are tropical agroforestry home gardens sustainable? Agr, Ecosyst.
19 Environ. 41: 189-207.
20 TROPEK, R. AND M. KONVICKA. 2010. Forest eternal? Endemic butterflies of the Bamenda
21 highlands, Cameroon, avoid close canopy forest. Afr. J. Ecol. 48: 428-437.
‐
21
1 VASCONCELOS, S., P. RODRIGUES, L. PALMA, L. F. MENDES, A. PALMINHA, L. CATARINO, AND
2 P. BEJA. 2015. Through the eye of a butterfly: Assessing biodiversity impacts of
3 cashew expansion in West Africa. Biol. Conserv. 191: 779-786.
4 VLASANEK, P., SAM, L. AND V. NOVOTNY. 2013. Dispersal of butterflies in a New Guinea
5 rainforest: using mark–recapture methods in a large, homogeneous habitat. Ecol. Ent.
6 38: 560-569.
7 WHITWORTH, A., J. VILLACAMPA, A. BROWN, R.P. HUARCAYA, R. DOWNIE AND R. MACLEOD
8 2016. Past Human Disturbance Effects upon Biodiversity are Greatest in the Canopy;
9 A Case Study on Rainforest Butterflies. PloS one, 11: 0150520.
22
1 TABLE 1. Characteristics of the six land-use categories, with mean tree density and species
2 richness per 1 ha plot.
Description Dominant tree species Mean tree density Mean tree sp. (± SEM) richness (± SEM)
Natural forest Uncultivated forest Apodytes dimidiate, Galiniera 258 ± 50 15 ± 1 dominated by saxifrage, Syzygium guineense indigenous trees Millettia ferruginea and Chionanthus mildbraedii
Timber Monoculture timber Pinus patula, Grevillea robusta 751 ± 254 1 ± 0 plantation plantations or Eucalyptus camaldulensis
SMCF Semi-managed coffee Croton macrostachyus, Albizia 136 ± 27 15 ± 1 forest: mixed indigenous gummifera, Ehreta cymosa and
shade trees managed to Cordia africana provide optimal
conditions for cultivation of Coffea arabica
Open woodland Patchy open woodland Maesa lanceolate and Acacia 122 ± 47 11 ± 1 abyssinica
Pasture Areas grazed by Acacia abyssinica and Ficus 10 ± 3 4 ± 1 livestock vasta
Cropland Cultivated for annual Cordia africana and Acacia 7 ± 2 2 ± 0.5 crops (maize, sorghum abyssinica and teff)
23
1 TABLE 2. Results from best-fitting linear mixed-effect models explaining butterfly
2 abundance and Hill’s diversity (0D, 1D, 2D). Models included survey round and altitudinal
3 zone as random effects. Marginal R2 values represent the variation explained by the
4 associated fixed effect, with the significance determined by comparing the fit of subsequent
5 models using Chi-squared. Asterisks indicate significance level (*** P< 0.001; ** P< 0.01).
Marginal AIC Δ χ2 2 R GLMM AIC
Abundance ~ land-use + distance from natural forest 0.51 377 0
~ land-use only 0.48 388 11 53.54 *** ~ distance from natural forest only 0.19 422 44 8.79 **
0D ~ land-use + distance from natural forest 0.49 207 0 ~ land-use only 0.48 218 11 53.04 *** ~ distance from natural forest only 0.14 274 68 9.10**
~ land-use + distance from natural forest 0.46 174 0 1D ~ land-use only 0.46 183 10 72.92 *** ~ distance from natural forest only 0.13 222 49 7.53 **
~ land-use + distance from natural forest 0.38 190 0
2D ~ land-use only 0.38 200 10 55.25 *** ~ distance from natural forest only 0.10 221 31 7.20 **
6
24
1 FIGURE 1. (a) Species accumulation curves and (b) Hill’s diversity associated with the six
2 land-use categories. Hill’s diversity indices represent the mean diversity per plot (± SEM)
3 and are weighted to the order of q, which reflects the sensitivity of the indices to the relative
4 abundance of species: q=0 is sensitive to rare species, q=1 is sensitive to common species and
5 q=2 is sensitive to highly abundant species.
6 FIGURE 2. Species similarity of butterfly communities in the agricultural land-use categories
7 as compared to natural forest. Species similarity is calculated using three indices that are
8 weighted to the order of q; q=0 represents similarity of rare species, q=1 of common species
9 and q=2 of abundant species.
10 FIGURE 3. Effect of distance from natural forest on butterfly abundance (a) and Hill’s
11 diversity (b)-(d).
12 FIGURE 4. Impact of land use on butterfly abundance across five ecological habitat
13 categories. Bars represent mean abundance per plot and error bars represent SEM.
25
1 FIG 1
2
26
1 FIG 2
2
3
27
1 FIG 3
2
28
1 FIG 4
2
29
1 Table S1. List of butterfly species and their total abundance within each land-use category.
2 Ecological habitat categories are defined according to Munyuli (2012): FDS= forest
3 dependent species, FEW= forest edge and woodland species, MS= migratory species, OHPS=
4 open habitat specialist, WSS= widespread species.
Habitat cat. NF SMCF PLNT WD CRP PAST
LYCAENIDAE
Polymmatinae 31 156 64 Polymmatinae sp1 41 195 58 Polymmatinae sp2
NYMPHALIDAE
Biblidinae WSS 3 11 1 4 Eurytela dryope (Cramer, 1775) Sevenia boisduvali FDS 1 22 3 (Wallengren, 1857)
Charaxinae FEW 1 5 5 3 1 Charaxes brutus (Cramer, 1779) Charaxes karkloof van Someren & Jackson, 1 1957
Danainae FDS 3 3 Amauris albimaculata Butler, 1875 2 17 4 1 Amauris echeria (Stoll, 1790) 5 10 1 Amauris ochlea (Boisduval, 1847) MS 6 7 4 1 1 3 Danaus chrysippus (Linnaeus, 1758) 5 6 Danainane sp1
Heliconiinae 16 4 6 1 1 1 Acraea acara Hewitson, 1865 9 3 Acraea aganice Hewitson, 1852 FDS 5 Acraea alciope Hewitson, 1852 5 2 Acraea anacreon Trimen, 1868 FEW 7 2 Acraea cabira Hopffer, 1855 1 Acraea cerasa Hewitson, 1861 WSS 5 5 Acraea encedon (L.) 26 13 Acraea esebria Hewitson, 1861 8 1 2 1 1 Acraea horta (L.) FDS 8 6 1 1 1 Acraea lycoa Godart, 1819 10 7 Acraea rahira Boisduval, 1833 WSS 20 5 3 Acraea serena (Fabricius, 1775)
30
MS 3 22 1 11 3 10 Phalanta phalantha (Drury, [1773])
Libytheinae MS 9 11 14 9 Libythea labdaca Westwood, 1851
Limenitidinae 18 30 1 2 5 Neptis goochii Trimen, 1879 WSS 33 67 12 10 4 Neptis laeta Overlaet, 1955 WSS 9 13 1 1 2 Neptis saclava Boisduval, 1833 FDS 2 Pseudacraea eurytus (L.)
Nymphalinae 7 13 1 2 4 3 Hypolimnas anthedon (Doubleday, 1845) MS 6 10 1 2 Hypolimnas misippus (Linnaeus, 1764) MS 1 Junonia hierta (Fabricius, 1798) 2 Junonia natalica (Felder & Felder, 1860) WSS 3 8 25 1 Junonia oenone (Linnaeus, 1758) WSS 13 1 11 9 4 Junonia terea (Drury, 1773) WSS 10 Precis octavia (Cramer, 1777) FDS 5 12 10 1 Protogoniomorpha anacardii (L.) FDS 3 9 13 Protogoniomorpha parhassus (Drury, 1782) 6 14 2 1 Vanessa dimorphica (Howarth, 1966)
Satyrinae OHPS 35 54 66 18 1 Bicyclus anynana (Butler, 1879) 51 145 60 12 6 Bicyclus safitza (Westwood, 1850) 9 17 12 4 4 1 Satyrinae sp1 WSS 2 11 2 5 13 Melanitis leda (Linnaeus, 1758) OHPS 9 39 21 9 11 Ypthima asterope (Klug, 1832) OHPS 15 40 12 11 16 3 Ypthima impure Elwes & Edwards, 189
PAPILIONIDAE
Papilioninae MS 2 2 3 1 Graphium Leonidas (Fabricius, 1793) WSS 28 20 4 Papilio dardanus Brown, 1776 MS 14 12 3 2 2 1 Papilio demodocus Esper, 1798 4 6 7 1 Papilio euphranor Trimen, 1868 FEW 11 19 11 3 1 2 Papilio nireus (L.)
PIERIDAE
Coliadinae MS 9 1 1 Catopsilia florella (Fabricius, 1775) 3 5 5 Catopsilia gorgophone (Boisduval, 1836) 7 10 1 3 2 5 Catopsilia sp1 MS 95 374 102 32 2 3 Colias electo (L.) 6 16 2 Coliadinae sp1
31
MS 8 10 14 3 Eurema brigitta (Stoll, [1780]) MS 17 13 7 Eurema desjardinsii (Boisduval, 1833) MS 54 43 31 6 2 6 Eurema hecabe (L.) 6 12 1 Coliadinae sp2
Pierinae MS 16 4 14 6 23 11 Belenois aurota (Fabricius, 1793) MS 13 15 4 5 2 11 Belenois creona (Cramer, 1776) MS 14 6 2 13 19 12 Belenois gidica (Godart, [1819]) FDS 6 9 Belenois raffrayi (Oberthür, 1878) 21 14 11 7 10 16 Belenois zochalia (Boisduval, 1836) WSS 3 13 8 1 Leptosia alcesta (Stoll, [1782]) 9 59 8 7 8 9 Mylothris rueppellii (Koch, 1865) 22 46 4 1 3 3 Pieris brassicae (L.) 6 19 1 1 1 Pieris rapae (L.)
1
32
1 Table S2. The effect of land-use on butterfly abundance within common sub-families. Results
2 from linear mixed-effect models including survey month and altitudinal zone as random
3 effects. Marginal R2 values represent the variation explained by land-use, with the
4 significance determined by comparing the fit of land-use model to a null model using Chi-
5 squared. Asterisks indicate significance level (*** P< 0.001; ** P< 0.01).
Marginal AIC Delta AIC χ2 (df=5) 2 R GLMM
Pieridae Coliadinae ~ Land-use 0.54 381 0 107.47*** ~ Null model 472 91 Pierinae ~ Land-use 0.17 352 0 ~ Null model 366 14 31.21***
Nymphalydiae Satyrinane ~ Land-use 0.43 381 0 75.02*** ~ Null model 440 59 Heliconiinae ~ Land-use 0.20 357 0 31.26*** ~ Null model 370 13 Lycaenidae ~ Land-use 0.10 381 0 19.09** Lycaeninae ~ Null model 383 2 Papilionidae ~ Land-use 0.19 280 0 32.58*** Papilioninae ~ Null model 292 12
6
33
1 Table S3. The effect of land use on butterfly abundance within ecological habitat categories.
2 Results from linear mixed-effect models including survey month and altitudinal zone as
3 random effects. Marginal R2 values represent the variation explained by land-use, with the
4 significance determined by comparing the fit of land-use model to a null model using Chi-
5 squared. Asterisks indicate significance level (*** P< 0.001; ** P< 0.01).
Marginal AIC Delta AIC χ2 (df=5) 2 R GLMM
Ecological habitat category Forest ~ Land-use 0.22 593 0 23.44*** Dependent ~ Null model 609 15 Forest ~ Land-use 0.13 455 0 13.81* Edge/Woodland ~ Null model 456 1
Migratory ~ Land-use 0.29 1021 0 24.36** species ~ Null model 1056 35 Open habitat ~ Land-use 0.20 794 0 18.28** specialists ~ Null model 813 19 Widespread ~ Land-use 0.08 1231 0 0.013* ~ Null model 1259 28
6
34
1 Figure S1. Map of the study site in the Jimma Highlands, southwestern Ethiopia depicting
2 main land-use categories along the altitudinal transect. White circles indicate locations of
3 sampled plots.
4
5
6
7
8 Figure S2. Abundance of sub-families within the six land-use categories. Bars represent mean
9 abundance per plot and error bars represent SEM.
35
1
2 A Pieridae: Coliadinae B Pieridae: Pierinae 2 2 16 R GLMM= 0.54 7 R GLMM= 0.17 3 14 6 12 4 5 10 4 8 5 3 6 2 6 Abundance 4 2 1 7 0 0 8
9
10 C Nymphalidae: Satyrinae D Nymphalidae: Heliconiinae 11 2 2 16 R GLMM= 0.43 8 R GLMM= 0.20 12 14 7 12 6 13 10 5 8 4 14 6 3
15 Abundance 4 2 2 1 16 0 0
17
18
19 E Lycaenidae: Lycaeninae F Papilionidae: Papillioninae 2 2 30 R GLMM= 0.10 4.5 R GLMM= 0.19 E Lycaenidae: Polymmatinae 4 25 2 2 R GLMM= 0.10 3.5 R GLMM= 0.15 20 3 2.5 15 2 10 1.5 1 Abundance 5 0.5 0 0
36
1 Figure S3. Non-metric multidimensional scaling plot illustrating butterfly community
2 structure in relation to land-use. Circles represent butterfly species, with colours indicating
3 their family. Land-use categories: F= natural forest, PLNT= plantation, SMCF= semi-
4 managed coffee forest, W= open woodland, C= cropland and P= pasture.
Lycaenidae Nymphalidae Papilionidae Pieridae
F
PLNT C SMCF
NMDS2 W P
-1.0 -0.5 0.0 0.5
-0.5 0.0 0.5 1.0
NMDS1
37